Reverse engineering synthetic antiviral amyloids
Emiel Michiels 1,2,10, Kenny Roose 3,4,10, Rodrigo Gallardo 1,2, Ladan Khodaparast1,2, Laleh Khodaparast1,2, Rob van der Kant 1,2, Maxime Siemons1,2,5, Bert Houben 1,2, Meine Ramakers1,2, Hannah Wilkinson1,2, Patricia Guerreiro1,2, Nikolaos Louros1,2, Suzanne J. F. Kaptein 6, Lorena Itatí Ibañez 3,4, Anouk Smet3,4, Pieter Baatsen1,7, Shu Liu8, Ina Vorberg 8,9, Guy Bormans5, Johan Neyts6, Xavier Saelens 3,4,
Frederic Rousseau 1,2✉& Joost Schymkowitz 1,2✉
Human amyloids have been shown to interact with viruses and interfere with viral replication.
Based on this observation, we employed a synthetic biology approach in which we engineered virus-specific amyloids against influenza A and Zika proteins. Each amyloid shares a homologous aggregation-prone fragment with a specific viral target protein. For influenza we demonstrate that a designer amyloid against PB2 accumulates in influenza A-infected tissue in vivo. Moreover, this amyloid acts specifically against influenza A and its common PB2 polymorphisms, but not influenza B, which lacks the homologous fragment. Our model amyloid demonstrates that the sequence specificity of amyloid interactions has the capacity to tune amyloid-virus interactions while allowing for the flexibility to maintain activity on evolutionary diverging variants.
https://doi.org/10.1038/s41467-020-16721-8 OPEN
1VIB Center for Brain and Disease Research, Leuven, Belgium.2Switch Laboratory, Department of Cellular and Molecular Medicine, KU Leuven, Leuven, Belgium.3VIB Center for Medical Biotechnology, Ghent, Belgium.4Department of Biomedical Molecular Biology, Ghent University, Ghent, Belgium.
5Laboratory for Radiopharmacy, Department of Pharmaceutical and Pharmacological Sciences, KU Leuven, Leuven, Belgium.6Laboratory of Virology and Chemotherapy, Rega Institute for Medical Research, Department of Microbiology, Immunology and Transplantation, KU Leuven, Leuven, Belgium.7Electron Microscopy Platform of VIB Bio Imaging Core, KU Leuven, Leuven, Belgium.8German Center for Neurodegenerative Diseases (DZNE e.V.), Bonn, Germany.
9Rheinische Friedrich-Wilhelms-Universität Bonn, Bonn, Germany.10These authors contributed equally: Emiel Michiels, Kenny Roose.✉email:frederic.
[email protected];[email protected]
1234567890():,;
F
or decades, human amyloids were mainly studied as the major hallmarks of protein deposition in disease, but in recent years functional amyloids, which have been uncov- ered in several human tissues1 and throughout all kingdoms of life, have caused a revision of the view that amyloid structures are intrinsically toxic or pathological. Recently, increasing evidence supports the hypothesis that even disease-associated amyloids may not only be pathogenic byproducts2, but that they could in fact also possess functional benefits for instance acting as inter- actors of human pathogens, including viruses3–6. For example, it has been suggested that the Alzheimer’s disease (AD)-associated peptide, amyloid beta (Aβ), may have a role in innate immunity against herpes virus brain infections5. More specifically, Eimer et al.5 showed that Aβ interacts with herpes glycoproteins, thereby inducing Aβ amyloid formation. The seeding of Aβ aggregation leads to an entrapment of the viral particles, which results in an amyloid-driven antiviral effect. Whether reciprocally the amyloid deposits formed as a result of such encounters with virus have a role in the development of AD remains to be determined, but herpes infection is a known risk factor for AD7. Aggregation seeding, as observed for Aβ when binding to herpes virus, is a well-described process, driven by so-called aggregation-prone regions (APRs)8. These APRs engage into homotypic interactions to form tightly packed intermolecular amyloid structures, resulting in a highly sequence-specific pro- cess9–12, but allowing the interaction between proteins sharing highly homologous APRs. Interestingly, Aβalso shares a homolog APR sequence with the envelope glycoprotein B4, a herpes pro- tein for which an interaction with Aβhas already been shown5. To study whether an amyloid that shares a homolog APR with a certain viral protein can indeed specifically interact with this protein and thereby interfere with viral replication, we resorted to synthetic biology approach.We designed two synthetic amyloids, each encoding a virus- specific APR identified in influenza A and Zika virus (ZIKV) proteins, respectively. We found that each amyloid interferes with the replication of its corresponding virus, without any cross- reactivity. For influenza A we show that our synthetic amyloid accumulates at the site of infection and interferes with influenza A replication in a murine infection model. The amyloid binds to its specific viral target protein in an APR-specific manner, forcing the protein into a nonfunctional conformation. Influenza B, which lacks the homolog APR, is not affected by the synthetic amyloid, highlighting the sequence specificity of this interaction.
Together, we present the design of fully synthetic amyloids, solely based on an APR identified in a viral protein, that show specific antiviral properties. Moreover, our results demonstrate that APR- mediated amyloid interactions possess the same characteristics defining bona fide functional biological interactions including specificity and in vivo selectivity.
Results
Design of an APR-based synthetic peptide with antiviral properties. To investigate the potential for amyloids to interact with viruses encoding a homologous APR, we here used a syn- thetic biology approach, using influenza A as a model virus. To design synthetic amyloids that share a homologous APR with an influenza A protein, wefirst identified all APRs in the influenza A proteome using the statistical thermodynamics algorithm TANGO13. We identified 40 APRs with a minimum TANGO score of 5 in the influenza A/Puerto Rico/8/34 (A/PR8) proteome.
Synthetic peptides were designed based on those APRs using a tandem repeat scaffold14–16: two identical APRs, supercharged with arginine residues and linked by two amino acids (Fig. 1a, Supplementary Table 1). The tandem design is used to stimulate
amyloid formation of our synthetic peptides, while at the same it increases avidity of APR interactions. The charged amino acids, on the other hand, are added to decrease the kinetics of self- aggregation. These amyloid-prone peptides were screened for antiviral activity against two influenza A strains: the H1N1 strain A/PR8 and the H5N1 strain A/NIBRG-14 (Supplementary Fig. 1a, b). Madin–Darby canine kidney (MDCK) cells were treated with peptide at a concentration of 10 µM, 2 h prior to infection. Following overnight incubation, the amount of newly produced virus was determined by TCID50 titration. Overall, peptide 12, derived from an APR identified in the cap-binding domain of polymerase basic protein 2 (PB2), 381LIQLIVS387
(Fig. 1b), showed the most pronounced effect. Importantly, the observed antiviral effect did not originate from aspecific cytotoxic effects (Supplementary Fig. 1c). We selected the APR of peptide 12,381LIQLIVS387, as the primary target APR and optimized the peptide design to maximize antiviral activity and minimize hemolytic activity. Different peptide variants were designed:
multiple charged amino acids and linker residues were screened and mutations were inserted inside the APR to increase solubility while retaining antiviral activity (Supplementary Table 2). The antiviral effect of all peptide variants was screened by a plaque- size reduction assay and a hemolytic activity assay was performed on human red blood cells (Supplementary Fig. 1d, e). Taking both parameters into account, the best scoring peptide sequence is WDLIQLIVSDGSDLIQLIVSD, referred to as peptide 12B, in which the target APR (381LIQLIVS387) isflanked by negatively charged aspartate residues and an N-terminal tryptophan is added to allow for accurate concentration determination. This plaque-size reduction assay showed an ~80% reduction of influ- enza A/PR8 plaque area when cells were treated with peptide 12B (Fig.1c, d). The antiviral effect was compared with three known anti-influenza compounds: nucleozin, a nucleoprotein-targeting inhibitor17, Tamiflu, a neuraminidase inhibitor, and VX-787, a PB2 inhibitor. Of note, positively charged variants of peptide 12B (12E–12H) show a similar antiviral effect in the plaque-size reduction assay but were not selected for further analysis because of the aspecific toxic effects on human red blood cells (Supple- mentary Fig. 1e). A dose–response experiment, in which influ- enza A/PR8-infected MDCK cells were treated with a peptide concentration ranging from 10 nM to 75 µM, allowed to deduce an IC50of ~1.79 µM for peptide 12B (Fig.1e). The cytotoxic dose (TD50=727 µM), derived from the hemolytic activity (Fig. 1e), and the IC50 allowed us to estimate a therapeutic index (TI= TD50/IC50) for peptide 12B of 406. Moreover, no aspecific toxic effects were observed on MDCK and Human Embryonic Kidney (HEK 293T) cells with concentrations up to 250 µM (Fig.1f and Supplementary Fig. 1f). Finally, we used two additional assays to validate the antiviral effect of peptide 12B: a multicycle replication assay and the protection against cytopathic effects (Supplemen- tary Fig. 1g, h). Both assays confirm that peptide 12B acts as an antiviral peptide against influenza A.
Peptide 12B shows antiviral activity in an in vivo infection model. To determine whether peptide 12B retains its antiviral activity in vivo, cohorts of BALB/c mice were infected with influenza A/PR8 (2x LD50) and treated daily with buffer, 2.5 mg/
kg peptide 12B (both intravenously (i.v.)), or 25 mg/kg Tamiflu (oral gavage). After 6 days, a small but significant reduction in virus titers in lung homogenates was observed for 12B-treated mice compared with buffer-treated and Tamiflu-treated mice18 (Fig.1g). To control for any possible aspecific toxic effects of our synthetic peptide, we performed an in vivo toxicity study. BALB/c mice received daily i.v. injections for 2 weeks with 10 mg/kg peptide 12B or buffer. Blood analysis and a histopathological
evaluation of all major organs revealed no abnormalities (Sup- plementary Fig. 2a–c and Supplementary Note 1). Biodistribution studies with a radiolabeled variant ([68Ga]Ga-NODAGA-PEG2- 12B) in influenza A/PR8-infected BALB/c mice demonstrated a lack of accumulation in healthy tissue, fast blood clearance, and a high urinary elimination without kidney retention (Supplemen- tary Fig. 2d, e). The radiolabeled peptide variant also allowed to estimate the half-life of peptide 12B (<10 min in blood), explaining the modest reduction in viral titers observed in vivo (Fig.1g) and underlines the need for more stable peptide variants
to obtain a more robust therapeutic activity. Moreover, uptake of radiolabeled peptide 12B was approximately twofold larger in influenza A-infected lungs compared with healthy lungs at dif- ferent time points after peptide injection, consistent with a spe- cific amyloid–virus interaction in vivo (Fig.1h).
Peptide 12B organizes into amyloid-like structures. We next performed biophysical studies of peptide 12B in vitro to verify that it behaves as an amyloid, in accordance with the design. Dynamic light scattering (DLS) showed that peptide 12B (100 µM) organizes
a b
TD50 = 727 ± 61.2 µM
IC50 = 1.79 ± 0.12 µM
c
e
0.001 0.01 0.1 1 10 100 0
20 40 60 80 100 120
0 20 40 60 80 100 120
Concentration 12B (µM)
Normalized area covered by plaques (%) Normalized RBC lysis (%)
1 10 100 1000 0
20 40 60 80 100 120
Concentration 12B (µM)
Normalized cell viability (%)
d
MDCK cells Red blood cells 90°
f
Target APR PB2CB
g h
DMSO
12B
(10 µM) Nucleozin (10 µM)
Tamiflu (100 µM)
Medium
VX-787 (1 µM) Charged amino acid
Linker APR
p < 0.0001 p < 0.0001
DMSO12B NucleozinTamifluVX-787 0
20 40 60 80 100 120
Normalized area covered by plaques (%)
p < 0.0001 p = 0.0054
p = 0.02 p = 0.0021
p = 0.0021
Buffer 12B 12Bpro2Tamiflu 0
5 10 15 20
PFU/mL (x 105)
p = 0.0028
Healthy lungs Infected lungs
30 min 60 min 0.0
0.5 1.0 1.5
SUV
p = 0.021
p = 0.0002
Fig. 1 APR-based design of a synthetic peptide with antiviral activity against influenza A. aGraphical illustration of the peptide tandem design with the target APRs capped by charged amino acids (arginine or aspartic acid) and linked by a glycine-serine or proline-proline linker.bCap-binding domain of the influenza A/PR8 polymerase basic protein 2 (PB2CB) (PDB=4ENF)26,27. The target APR,381LIQLIVS387, is surface exposed and highlighted in red.c Plaque-size reduction assay of MDCK cells treated with cell medium, 1% DMSO, 10µM peptide 12B, 10µM nucleozin, 100µM Tamiflu, or 1µM VX-787 2 h prior to infection with influenza A/PR8.dQuantification of the area covered by plaques inc. Data are normalized to medium-treated cells and mean ± SD is shown (n=4 independent experiments, statistics: one-way ANOVA with multiple comparison).eDose-dependent effect of peptide 12B on area covered by plaques (red curve, left axis) and on red blood cell (RBC) lysis (purple curve, right axis). For IC50: data are normalized to buffer-treated cells and the mean ± SD is shown (n=4 independent experiments). For toxic dose (TD50): data are normalized to buffer-treated (0% lysis) and 0.1% Triton-treated cells (100% lysis) and the mean ± SD is shown (n=3 independent experiments).fDose-dependent toxicity of peptide 12B, after 24-h incubation on MDCK cells. Data are normalized to buffer-treated (100% viability) and 0.1% Triton-treated cells (0% viability) and the mean ± SD is shown (n=3 independent experiments).gViral load in the lungs of influenza A/PR8-infected mice after daily injections of either buffer, 2.5 mg/kg peptide 12B, 2.5 mg/kg peptide 12Bpro2 (all i.v.), or 25 mg/kg Tamiflu (oral gavage) for 6 consecutive days. Data are expressed as mean ± SD (n=13 mice over six independent experiments, statistics: one-way ANOVA with multiple comparison).hRelative concentrations (SUV) of peptide [68Ga]Ga-NODAGA-PEG2-12B in the lungs of A/PR8-infected versus healthy mice at different time points after peptide injection. The data are expressed as mean ± SD (n=12 mice over six independent experiments, statistics: two-sided unpaired Student’sttest).
into small structures (<50 nm) that evolve into larger aggregates (~1 µm) within 20 h (Fig.2a). Freshly dissolved peptide 12B already binds the amyloid-specific dye thioflavin-T (Th-T) and the Th-T spectra do not change significantly over time (Fig. 2b, c). The Fourier-transform infrared (FTIR) spectrum of a freshly dissolved peptide shows a main peak at 1624 cm−1, which is indicative of intermolecular beta-sheet formation and is hence typically observed for amyloid-like aggregates19(Fig.3d). To evaluate peptide aggre- gate morphology, we used transmission electron microscopy (TEM), showing amyloid-like structures of ±8–10 nm in width (Fig.3e). However, in contrast to typical amyloidfibers, peptide 12B appears to organize in shorter, curved fibers, sometimes forming annular aggregates20,21. Cryo-TEM images confirm that these structures are present in solution and are not an artifact of the drying or staining steps in regular TEM (Fig.3f). Together, these data indicate that peptide 12B organizes into small beta-sheet- containing structures that grow into larger amyloid-like aggregates
over time. Importantly, DLS results indicate that these small oli- gomeric structures can be stabilized by simply keeping peptide 12B at a low concentration (10 µM), which is approximately fivefold higher than its IC50, as described in the antiviral assays above (Fig.3g). Finally, these small oligomeric structures are soluble, as shown by ultracentrifugation followed by concentration determi- nation at different time points (Fig.3h).
Forming amyloid is necessary but insufficient for antiviral activity. As controls, we designed two additional peptides, in which we maintained an identical scaffold but altered the APR cores (Supplementary Table 3). Peptide 12Bpro2 was designed as a control for the role of beta-strand-mediated aggregation by introducing two proline mutations in each APR to reduce the aggregation propensity while avoiding the introduction of additional charged residues22. Peptide 12Bscr2 was designed as a control for the sequence specificity of aggregation by
0 5 10 15 20
0 5000 10,000 15,000 20,000
Time (h) ThT fluorescence at 485 nm (a.u.)
1500 1600
1700 0 10 20 30
wavenumber (nm–1)
Absorbance units
1624
1688 1650
500 550 600
0 5000 10,000 15,000
wavelength (nm)
ThT fluorescence (a.u.)
Light scattering (100 µM) Amyloid-specific dye (100 µM)
Electron microscopy (100 µM) FTIR (100 µM)
Light scattering (10 µM)
0 5 10 15
0 20 40 60 80 100
Time (h)
0 5 10 15 20
0 250 500 750 1000
Time (h) Average RH (nm)Average RH (nm)
Cryo-electron microscopy (100 µM) 12B
200nm 8nm 40nm
100nm Amyloid-specific dye (100 µM)
d e f
g h i
Solubility
0 1 2 3 4 5 6 7 0
20 40 60 80 100 120
Time (h)
Soluble fraction
a b c
Buffer12B 12Bpro212Bscr2 0
50 100 150
Normalized area covered by plaques (%)
p < 0.0001
12B
12Bpro2 12Bscr2 Peptide core sequence
Fig. 2 Peptide 12B organizes into amyloid-like structures in vitro. aHydrodynamic radius (RH) calculated from the regularizationfit of DLS data of peptide 12B (100µM) maturation over time.bTh-T emission spectrum after excitation at 440 nm of buffer (black line) or peptide 12B (100µM, purple line) 5 min after solubilization (n=3 independent experiments).cTh-T emission signal of buffer (black line) or peptide 12B (100µM, purple line) at 485 nm after excitation at 440 nm as a function of time. Data represent three independent repeats (n=3 independent experiments).dFTIR spectrum of 100µM peptide 12B 5 min after solubilization. TEM (e) and cryo-TEM (f) of 100µM freshly dissolved peptide 12B (both are one representative image of three independent experiments).gHydrodynamic radius (RH) calculated from the regularizationfit of DLS data of peptide 12B (10µM) maturation over time (dashed line= 50 nm).hSoluble fraction of peptide 12B (10µM) determined after ultracentrifugation (250,000 ×gfor 30 min), measured over time, and mean ± SD is shown (n=3 independent experiments).iQuantification of the area covered by plaques of MDCK cells treated with 10µM peptide in a plaque-size reduction assay with influenza A/PR8. Data are normalized to medium-treated cells and the mean ± SD is shown (n=5 independent experiments, statistics: one-way ANOVA with multiple comparison).
scrambling the amino acids of the APR while maintaining its propensity to form amyloids. As intended by their design, an in vitro biophysical analysis showed that, in our experimental conditions, 12Bscr2 (100 µM) organizes into amyloid-like aggregates over time, while 12Bpro2 (100 µM) does not aggre- gate (Supplementary Fig. 3a–f). Only at a higher concentra tion (500 µM), peptide 12Bpro2 organizes into amorphous
aggregates (Supplementary Fig. 3d). Treatment of influenza A/
PR8-infected MDCK cells with these control peptides showed no significant effect on viral replication, indicating that peptide amyloid formation is necessary but in itself not sufficient for antiviral activity (Fig.2i, Supplementary Fig. 3g, h). Moreover, peptide 12Bpro2 did not show inhibitory effects on influenza A replication in the murine infection model (Fig. 1g).
0 2 4 6
0 40,000 80,000 120,000
Time (h)
Scattering at 266 nm (a.u.)
PB2CB PB2CB + 12B
PB2CB + 12Bpro2 PB2CB + 12Bscr2
0 250 500 750
-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2
Time (s)
Responce (RU)
12B 12Bpro2
12Bscr2
28 PM 3.5 PM 0.11 PM 0 PM
m7GTP- atto488 + PB2CB
IC50 = 3.52 ± 0.26 PM
0.001 0.01 0.1 1 10 100 0
20 40 60 80 100
Concentration 12B (PM)
Normalized FP after 20 h
0 5 10 15
0 10,000 20,000 30,000 40,000
Time (h)
ThT fluorescence (a.u.)
0 5 10 15
0 10,000 20,000 30,000
Time (h)
ANS fluorescence (a.u.)
40 nm
PB2CB + 12B
Association Dissociation
a b
c
d
e
f
g h
1 Pm
0 10,000 20,000 30,000
40,000 p < 0.0001
ThT fluorescence at 15 h (a.u.)
0 10,000 20,000 30,000
ANS fluorescence at 15 h (a.u.)
p < 0.0001
Buffer12B 12Bpro212Bscr2 0
20 40 60 80 100
Normalised FP after 20 h
p < 0.0001 p = 0.021 p < 0.0001 p < 0.0001
0 20,000 40,000 60,000 80,000
100,000 p < 0.0001
Scattering at 266 nm at 15 h (a.u.)
p = 0.023
0 5 10 15 20 25
20 60 100 140
Time (h)
FP (mP)
Amyloid peptide 12B interacts with influenza via PB2 in vitro.
To provide mechanistic understanding of the amyloid–virus interaction, we used recombinant expression and purification of the PB2 cap-binding domain of influenza A/PR8 (PB2CB), the intended target protein of peptide 12B (Supplementary Fig. 4a), and performed direct interaction measurements between the synthetic amyloid 12B and the purified protein. Biolayer inter- ferometry (BLI) showed a clear binding event of amyloid peptides 12B and 12Bscr2 to immobilized PB2CB, while 12Bpro2 did not show any affinity for PB2CB(Fig.3a). The dissociation constant (Kd) of peptide 12B to PB2CB(65.8 ± 0.91 nM) is >10-fold lower compared with 12Bscr2 (714 ± 33.9 nM) (Supplementary Fig. 4b, c). PB2CBshowed an increased binding of 8-anilinonaphtalene-1- sulfonic acid (ANS) when seeded with amyloid peptide 12B, indicative of a conformational change of the protein that leads to the exposure of hydrophobic residues normally buried in the folded protein (Fig.3b). In agreement with the latter observation, 12B, and not control peptides 12Bpro2 and 12Bscr2, increased Th-T binding and scattering intensity of PB2CB, consistent with beta aggregation (Fig. 3c, d). Importantly, this increase in ANS fluorescence, Th-Tfluorescence, and scattering did not originate from the aggregation of the peptides alone (Supplementary Fig. 4d–f). In addition, TEM images showed that the final structures formed by 12B-induced PB2CBaggregates arefiber-like (Fig. 3e), compared with 12Bpro2-, 12Bscr2-, or noninduced PB2CB aggregates, which have an amorphous structure (Supple- mentary Fig. 4g). All together, these data show that peptide 12B binds with nanomolar affinity to PB2CB, inducing a conforma- tional change of the natively folded PB2CB, resulting in the for- mation of beta-aggregates. To assess that this leads to loss of function, we studied the so-called“cap-snatching”process, which occurs at the initiation of viral RNA synthesis, in which PB2 binds host pre-mRNA via a 5′cap structure (m7GTP). To validate whether specific 12B-induced aggregation is associated with a loss of PB2 cap-binding function, we used afluorescent polarization assay (Supplementary Fig. 4h)23. When a fluorescently labeled ligand (m7GTP-atto488) and purified PB2CB were mixed with peptide 12B, a concentration-dependent loss of cap-binding function was observed, which further decreased over time (Fig.3f, g; Supplementary Fig. 4i). Control peptide 12Bpro2 had no effect on PB2 cap-binding activity, while 12Bscr2 showed only a mild effect, possibly due to the modest affinity of this peptide to PB2CB
(Fig.3h).
Amyloid peptide 12B interacts with influenza via PB2 in cel- lulo. The in vitro results described in the previous section show that amyloid peptide 12B binds to PB2, induces its aggregation, and interferes with its function. Next, we studied whether the same event occurs in the cellular cytoplasm. To establish that in
a cellular context the synthetic amyloid does not interfere with viral entry in an aspecific manner, we performed a time of addition assay. This assay shows that the antiviral activity of 12B is unchanged, whether it is used in a pretreatment setup or added 6 h after infection (Fig. 4a, Supplementary Fig. 5a).
Moreover, when influenza A virion particles are pretreated with amyloid peptide 12B, the antiviral activity does not change dramatically compared with pretreatment of the cells (Fig.4b, Supplementary Figs. 1e and 5b). All the later indicates that 12B does not aspecifically interfere with viral entry and suggests that the amyloid–virus interaction occurs in the cell. To study the interactions between amyloid peptide 12B and influenza A in the cell and to confirm that this interaction also induces aggregation and inactivation of PB2, we used a modified influenza A/WSN/1933(H1N1) strain, with a FLAG-tagged PB2 (A/WSN-FLAG), facilitating PB2 detection. Immuno- fluorescence experiments showed co-localization of a FITC- labeled peptide 12B variant with PB2 (Fig.4c). To support the latter observation, we used a PEG2-biotin labeled variant of peptide 12B, 12Bpro2, and 12Bscr2 to perform a co- immunoprecipitation of PB2 from lysates of peptide-treated, A/WSN-FLAG-infected MDCK cells. The enrichment of PB2 in the co-immunoprecipitated fraction of 12B-PEG2-biotin-trea- ted MDCK cells is consistent with a direct interaction between amyloid peptide 12B and the target protein PB2 (Fig. 4d).
Finally, peptide 12B treatment of A/WSN-FLAG-infected MDCK cells resulted in a reduction of soluble PB2, while the control peptides had no effect (Fig.4e), showing PB2 undergoes aggregation. Taken together with the significant reduction in viral replication upon treatment with 12B shown above (Fig.1c, d), these data confirm that amyloid peptide 12B interacts with PB2 in the cytoplasm of influenza A-infected MDCK cells and induces its inactivation through aggregation.
Amyloid peptide 12B acts in an APR-specific manner.
According to our model, the underlying properties of amyloid interactions confer specificity upon the amyloid–virus interaction, but it was unclear how much tolerance to sequence variation is present. An in silico analysis of over 40,000 influenza A strains showed that the target APR in PB2 is highly conserved (Supple- mentary Fig. 6a). We evaluated the antiviral activity of amyloid peptide 12B on a set of influenza A strains, across different sub- types (H1N1, H3N2, H3N3, H3N8, H5N1), in which the fre- quently occurring single polymorphisms of the target APR are represented (381LIQLIVS387to381LVQLIVS387or381LFQLIVS387).
Peptide 12B retains activity against all these influenza A strains, suggesting the interaction allows for some tolerance (Fig.5a). In addition, we performed an in silico mutational analysis in which the effect of two mutations in the APR on the interaction potential
Fig. 3 Amyloid peptide 12B interacts with influenza via the PB2 target protein in vitro. aBiolayer interferometry showing association and dissociation of 10µM peptide to his-tagged, immobilized PBCB.b,cANS (excitation 380 nm, emission 485 nm) and Th-T (excitation 440 nm, emission 485 nm) fluorescence of PB2CB(32µM) with and without peptides (3.2µM) over time. Quantification of thefluorescence signal after 15 h is shown on the right.
Data are expressed as mean ± SD (n=3 (b) andn=6 (c) independent experiments, statistics: one-way ANOVA with multiple comparison to PB2CB
alone).dParticle scattering (260 nm) of PB2CB(32µM) with and without peptides (3.2µM) over time. Quantification of scattering signal after 1 h (before precipitation) is shown on the right. Data are expressed as mean ± SD (n=4 independent experiments, statistics: one-way ANOVA with multiple comparison to PB2CBalone).eTEM images of PB2CB(32µM) with peptide 12B (3.2µM) after 15 h incubation.fBinding of m7GTP-atto488 by PB2CB
in absence or presence (gray scale for peptide concentration) of a twofold serial dilution of peptide 12B, determined byfluorescence polarization (FP).
Mean ± SD is shown (n=4 independent experiments). The dashed lines represent maximal binding of m7GTP-atto488 to PB2 (upper dashed line) and free m7GTP-atto488 (lower dashed line), based on control experiments (Supplementary Fig. 4h).gConcentration-dependent effect of peptide 12B on the binding of m7GTP-atto488 by PB2CB(5µM) after 20 h incubation as shown by FP. Data represent normalized values (maximal binding=100% and free m7GTP-atto488=0%) and mean ± SD is shown (n=5 independent experiments).hBinding of m7GTP-atto488 by PB2CB(5µM) with or without peptides (7µM) after 20 h incubation. Data represent normalized values (similar as ing) and mean ± SD is shown (n=3 independent experiments, statistics: one- way ANOVA with multiple comparison).
was calculated (Supplementary Fig. 7, Supplementary Note 2).
Only 17.3% of all possible double-mutant combinations are pre- dicted to still interact with the381LIQLIVS387sequence, of which
>90% would have a strong negative effect on PB2 protein stability, and hence do not occur in natural influenza A strains. On the other hand, 12B did not interfere with the replication of several influenza B virus strains (B/Memphis/12/97 (B/Mem), B/Wis- consin/01/2010 (B/Wis), and B/Brisbane/60/2008 (B/Bris)), in which all have ~40% overall sequence homology to influenza A/
PR8 but lack the specific target APR, in vitro (Fig.5a, b). Of note,
the corresponding target sequence in influenza B (383MEKL- LIN389) is located in the same position in the cap-binding domain of PB2 compared with the target sequence (381LIQLIVS387) in influenza A PB2. In addition to the plaque-size reduction assay, influenza B replication is also unaffected in a multicycle replica- tion assay and 12B shows no protection against cytopathic effects of influenza B (Supplementary Fig. 6b, c). Of note, time of addi- tion assays and pretreatment of influenza B virion particles showed no antiviral activity of amyloid peptide 12B, again con- firming that our amyloid does not act as a nonspecific inhibitor of
DAPI Peptide PB2-FLAG
12B-FITC
12Bpro2-FITC
Buffer
MERGED
15 Pm 15 Pm
15 Pm
PB2 Vimentin GAPDH
S I S I S I S I S I
100 kDa 70 kDa
35 kDa 55 kDa
12B 12Bpro2 12Bscr2 Medium
Input CO-IP (streptavidin)
12B 12Bpro212Bscr2 100 kDa
15 kDa
PB2 Streptavidin monomer 70 kDa
12B Buffer
Buffer
Buffer 12Bpro212Bscr2
a c
IC50 = 0.93 ± 0.10 PM
0.01 0.1 1 10
0 20 40 60 80 100 120
Concentration (PM) Normalized area covered by plaques (%)
Virion pretreatment assay b
d
e
-4 -2 0 2 4 6
0 20 40 60 80 100 120
Time of treatment relative to infection (h) Normalized area covered by plaques (%)
Time of addition assay
Buffer 12B 12Bpro2 12Bscr2 0 10 20 30 40 50
% IP (relative to input)
p < 0.0001 p < 0.0001
p = 0.0018
Buffer 12B 12Bpro2 12Bscr2 0 50 100 150
Normalised soluble PB2 (%)
p = 0.019 p = 0.0014 p = 0.0004
Fig. 4 Amyloid peptide 12B interacts with influenza via the PB2 target protein in cellulo. aQuantification of the area covered by plaques of A/PR8- infected MDCK cells treated with 10µM peptide 12B at different time points relative to infection. Data are normalized to medium-treated cells and mean ± SD is shown (n=8 from three independent experiments, statistics: one-way ANOVA with multiple comparison,pvalue=0.0605). Representative images are shown in Supplementary Fig. 5a.bQuantification of the area covered by plaques in a plaque-size reduction assay of MDCK cells infected with A/PR8 virion particles that were pretreated with peptide 12B. Concentration-dependent effect of peptide 12B is shown normalized to medium-treated virion particles as mean ± SD (n=3 independent experiments). Representative images are shown in Supplementary Fig. 5b.cMDCK cells treated with FITC- labeled peptide 12B, 12Bpro2 (10µM), or buffer for 2 h prior to A/WSN-FLAG infection (MOI=1, 16-h infection). Cells werefixed and stained with an anti- FLAG antibody and DAPI. One representative example of three independent experiments.dWestern blot analysis of total (input) and immunoprecipitated fraction of MDCK cells treated with PEG2-biotin-labeled peptide (10µM) or buffer for 2 h prior to A/WSN-FLAG infection (MOI=1, 16-h infection).
Quantification of immunoprecipitated fraction of PB2, relative to PB2 input levels, is shown on the right. Data represent mean ± SD (n=5 independent experiments, statistics: one-way ANOVA with multiple comparison). Streptavidin was used as a loading control.eWestern blot analysis of PB2 distribution in soluble and insoluble fraction of lysates of MDCK cells treated with peptide (10µM) for 2 h prior to influenza A/WSN-FLAG infection (MOI=1, 16-h infection). Vimentin and GAPDH were used as loading controls for insoluble and soluble fraction, respectively. Quantification represents soluble PB2 fraction, normalized to medium-treated cells (100% soluble PB2). Data represent mean ± SD (n=4 independent experiments, statistics: one-way ANOVA with multiple comparison). Fordandethe samples on one blot are derived from the same experiment and the gels/blots were processed in parallel. Full blots are shown in Supplementary Fig. 12.
viral entry (Supplementary Fig. 5a, b). Influenza B replication is also not affected by peptide 12B in the murine infection model, confirming the APR-specific activity in vivo (Fig.5c). As expected from the lack of target APR in the protein, peptide 12B showed no effect on the aggregation or conformational stability of recombi- nantly purified PB2CB from influenza B/Mem (PB2IBV) as seen from Th-T and ANS binding assays (Fig.5d, e). In the same line, peptide 12B was not able to affect the aggregation of the highly
aggregation-prone amyloids when they share no similarity with peptide 12B (Aβ42 or amylin (IAPP), Fig.5f, g, respectively). To explore the limits of the sequence-identity tolerance, we searched for a naturally occurring aggregation-prone protein with imperfect identity to peptide 12B. The C-terminal domain (glycosylpho- sphatidylinositol-anchor sequence) of the prion protein (PrP) contains a region with high sequence homology to peptide 12B, both in human and mouse, covering only two mismatches (Q249F
A/PR8 B/Mem
374 376
396 398
0.0 0.2 0.4 0.6 0.8 1.0
0 500 1000 1500 2000 2500
0 1 2 3 4 5
0 500 1000 1500 2000 2500
Time (h)
ThT fluorescence (480 nm)
0 1 2 3 4 5
0 20,000 40,000 60,000 80,000
Time (h)
ANS fluorescence (a.u.) 0.0 0.2 0.4 0.6 0.8 1.0
0 20000 40000 60000 80000
PB2IBV PB2IBV + 12B
PB2IBV + 12Bpro2 PB2IBV + 12Bscr2
PB2IBV PB2IBV + 12B
PB2IBV + 12Bpro2 PB2IBV + 12Bscr2
Amyloid beta
Amyloid beta + Amyloid beta seeds
Amyloid beta + 12B
0 2 4 6 8 10
1500 2000 2500 3000 3500 4000
Time (h)
ThT fluorescence (a.u.)
IAPP
IAPP + IAPP seeds
IAPP + 12B
0 1 2 3
2000 4000 6000 8000
Time (h)
ThT fluorescence (a.u.)
12B APR 1
PrPHuman 247
6 252
22L Mock 12B 12Bpro2 12Bscr2
Total PrP
PrPSc
GAPDH
55 35 15 35 15 35 PM LF MW(kDa) - - + - + - + - + - + - + -
10 1 10 1 10 1
55 35 15 35 15 Total PrP 35
PrPSc
GAPDH
22L Mock 12B 12Bpro2 12Bscr2
- - + - + - + -
10 PM
LF
10 10
MW(kDa) Viral target
sequence H1N1H3N3H5N1H1N1 H3N2H3N2 H3N8
a b
c
d
e
f
i
g
h
j
A/PR8A/SwOA/NIB
A/WSN(F)A/UdorA/ChicA/DuckB/MemB/WisB/Bris 0
20 40 60 80 100 120 140 160
Normalized area covered by plaques (%) p = 0.0005 p = 0.0032 p = 0.0002 p = 0.0025 p = 0.0202 p = 0.0008 p = 0.0015
Buffer12B 12Bpro2Tamiflu 0
10 20 30 40
PFU/mL(x102)
0 500 1000 1500 2000 2500
ThT fluorescence after 15 h (a.u.)
0 20,000 40,000 60,000 80,000
ANS fluorescence after 15h (a.u.)
0 1000 2000 3000 4000
ThT fluorescence after 10h (a.u.)
p < 0.0001
0 2000 4000 6000 8000
ThT fluorescence after 3.5 h (a.u.)
p < 0.0001
and S253G, Fig. 5h). However, peptide 12B was incapable to convert PrP to the aggregated prion state (PrPSC) in cells expressing soluble mouse PrPC(Fig.5i, j). The combination of a sequence-specific process (not affecting influenza B or PrP) that is tolerant to single conservative substitutions (affecting the influ- enza A variants) suggests that the approach has potential as a broad-spectrum antiviral strategy to target influenza A.
Exploring the chemical space of antiviral synthetic amyloids. In order to get an indication if our approach with Influenza A is generally transferable, we set out to identify amyloids that target the ZIKV specifically. Following the same approach as described earlier, 63 different peptides were designed based on APRs identified in the ZIKV proteome (African strain MR766, Sup- plementary Table 4). Vero E6 cells were treated with peptide at a concentration of 20 µM, 2 h prior to infection. Following infec- tion for 48 h, the number of infected cells was determined by immunofluorescence. Multiple peptides showed inhibitory effects on ZIKV replication and protective effects toward the infected mammalian cell line (Supplementary Figs. 8 and 9). Overall, peptide R50 (RVAIAWLLRGSRVAIAWLLR), derived from an APR identified in membrane glycoprotein M, 140VAIAWLL146, showed the most pronounced effect compared with our positive control compound 7-DMA (7-deaza-2′-C-methyladenosine)24. A biophysical analysis confirmed the amyloid propensity of peptide R50. DLS showed that peptide R50 organizes into small structures (<50 nm) that evolve into larger aggregates (~300 µm) within 5 h (Fig. 6a). These aggregates are positive for the amyloid-specific dye Th-T (Fig.6b) and have an amyloid-like morphology as seen from TEM images (Fig. 6c). Even though peptide R50 organizes into amyloid structures, ~80% of total peptide remains in solution over time (Fig.6d). A dose–response experiment, in which Vero E6 cells were treated with different concentrations of peptide or compound, prior to viral infection showed IC50values of 5.25 ± 0.30 and 4.74 ± 0.31 µM for peptide R50 and 7-DMA, respectively (Fig. 6e, f, Supplementary Fig. 10a, b). Cross-validation with peptide 12B confirmed the APR-specificity associated with our approach as no interference with ZIKV replication is observed with this anti-influenza peptide (Fig.6g, Supplementary Fig. 10c).
Importantly, the concentration-dependent inhibitory effect of amyloid peptide R50 and 7-DMA on ZIKV replication did not originate from aspecific toxicity on the mammalian host cell (Fig.6h–j). Finally, reminiscent of amyloid peptide 12B, a time of addition assay, shows that R50 remains active, even when added 6 h after ZIKV infection (Supplementary Fig. 11a, b).
From these data it appears that our design approach is generally applicable and can be used to specifically target different viruses. In other words, these family of antiviral peptides
constitutes a novel chemical space that could be mined for novel antivirals that act through the mechanism of targeted protein aggregation.
Discussion
Human encoded amyloids possess functional roles and are not just pathogenic byproducts in disease1. The AD-associated amyloid, Aβ, physically interacts with herpes virus particles and interferes with viral replication. Building on this observation, we present here the reverse engineering of synthetic amyloids that specifically interact with a predefined target virus and reduce viral replication. Since it is well established that amyloid interactions are driven by the sequence-specific assembly of homologous aggregation-prone fragments8–12, we designed our amyloids by encoding a unique amino acid fragment corresponding to a viral APR. Using influenza A as a model virus, we designed an amyloid with a homologous APR to the PB2 protein. Our results show that this homologous APR drives the interaction between the amyloid and PB2, leading to aggregation and loss-of-function of the PB2 protein. We show that our amyloid does not aspecifically interfere with virus entry, instead amyloid–virus interaction occurs in the cytosol of the infected cell. Following amyloid uptake25, a sequence-specific binding event induces PB2 aggre- gation in the cytosol. Since APRs are usually buried inside the core of the folded protein, we suspect that this interaction hap- pens during protein translation15, when the APR might still be solvent exposed. PB2 loses its essential cap-snatching function following the amyloid-induced aggregation, leading to reduced viral replication. Importantly, our amyloid interacts with PB2 in a sequence-specific manner, since two mutations are sufficient to abrogate the interaction. Finally, the universal mechanism underlying homotypic APR interactions allowed us to extrapolate this technology resulting in a second amyloid that inhibits ZIKV replication.
Aβ encodes an APR matching an APR in envelope glycopro- tein B, a herpes protein4. Our results indicate that such homolog APRs are sufficient to drive the amyloid–virus interaction and provide a mechanistic understanding how such interactions would arise. Whether the suggested mechanism of interaction is broadly used for amyloid–virus interactions like the Aβamyloid and the herpes glycoprotein B remains to be determined together with the general role of amyloids in innate immunity. However, our work demonstrates that the increasing understanding of the functional aspects of amyloids could extend beyond purely amyloid-specific structural roles such as scaffolding or con- formational switches1,26–28. Instead, amyloid structures also entail the ability to engage in very specific interactions with other proteins under complex physiological conditions and affect their
Fig. 5 Amyloid peptide 12B acts as a sequence-specific interferor of influenza A. aQuantification of area covered by plaques of influenza-infected MDCK cells, treated with 10µM peptide 12B. Data are normalized to buffer-treated cells and mean ± SD is shown (n=4 independent experiments, statistics: two- sided one-samplet-test). A list of abbreviations is provided in“Methods.”bLocal alignment of PB2 from influenza A/PR8 and B/Mem. Target APRs are highlighted in red (A/PR8) and blue (B/Mem).cViral load in the lungs of influenza B/Mem-infected mice after daily injections of buffer, 2.5 mg/kg peptide 12B, 2.5 mg/kg peptide 12Bpro2 (i.v.); or 25 mg/kg Tamiflu (oral gavage) for 6 days. Data represent mean ± SD (n=13 mice,five independent experiments, statistics: one-way ANOVA with multiple comparison,pvalue=0.6027).d,eTh-T and ANSfluorescence of PB2CB(IBV) (32µM) without and with peptides (3.2µM, enlarged in the inserts). Fluorescence after 20 h is shown on the right as mean ± SD (n=3 independent experiments, statistics: one-way ANOVA with multiple comparison,pvalue=0.4218 fordand 0.1499 fore).f,gTh-Tfluorescence of amyloid beta (10µM) and IAPP (10µM) without and with peptide 12B (1µM) is shown. Self-seeded control is included by using preformed aggregates of amyloid beta and IAPP, respectively (1µM). Th-T fluorescence at plateau is shown on the right as mean ± SD (n=4 independent experiments, statistics: one-way ANOVA with multiple comparison).
hLocal alignment of the target APR in peptide 12B with human prion protein (PrPhuman). Western blot analysis of proteinase K-treated lysates of L929 15.9 cells (six passages, (i)) and CAD5 cells (eight passages, (j)) post treatment with peptides (1 or 10µM) with or without Lipofectamine 2000 (LF). As controls, cells were exposed to brain homogenate from a terminally diseased mouse infected with scrapie strain 22L (PrPSc) or with normal (Mock) brain homogenate. GAPDH was used as loading control. The samples are derived from the same experiment and the gels/blots were processed in parallel. Full blots are shown in Supplementary Fig. 12.
biological function. Finally, the ability to reverse engineer syn- thetic amyloids with specific biological activity also offers a new toolkit for protein functional interference and possibly ther- apeutic intervention.
Methods
Reagents. All reagents and antibodies are listed in Supplementary Table 5.
Mouse models. All mouse experiments were conducted according to the National (Belgian Law 14/08/1986 and 22/12/2003, Belgian Royal Decree 06/04/2010) and European (EU Directives 2010/63/EU, 86/609/EEG) animal regulations. All pro- tocols were approved by the Institutional Ethics Committee of Ghent University (Eth. Com. no. 2018-010). Five to eight weeks old healthy female wild-type BALB/
C mice were used in all mice experiments. Three or four mice were housed together in standardized cages with unrestricted access to food and water.
Tissue culture. MDCK cells were obtained from the American Type Culture Collection (ATCC). MDCK cells were maintained at 37 °C with 5% CO2 in DMEM medium (GIBCO), supplemented with 10% FCS (GIBCO),L-glutamine (GIBCO), and nonessential amino acids (GIBCO). HEK 293T cells were obtained from the
ATCC and maintained in DMEM medium, supplemented with 10% FBS, 1 mM sodium pyruvate, and nonessential amino acids. Mousefibroblast cell line L929 (ECACC; L929 (NCTC); #85103115) was purchased from Sigma-Aldrich (Tauf- kirchen, Germany). The L929 subclone 15.9, which is highly susceptible to the mouse-adapted scrapie strain 22L, was used for all experiments29,30. CAD5 cells31 were a kind gift of Corinne Lasmezas (The Scripps Research Institute, FL).
Bioinformatics. The statistical thermodynamics algorithm, TANGO, was used for all APR identifications. A cutoff on the TANGO score offive per residue was selected, which results in a Matthews correlation coefficient between prediction and experiment of 0.92. Settings within TANGO: temperature=298 K; pH=7.5; ionic strength=0.15 M; and minimum length of APRs=5.
Viruses. Viruses used throughout the work include A/Puerto Rico/8/34 (H1N1):
A/PR8, A/NIBRG-14 (H5N1): A/NIB, A/Swine Ontario/42729A/2001 (H3N3): A/
SwO, A/Udorn/307/72 (H3N2): A/Udor, A/Chicken/Nanchang/3-120/2001 (H3N2): A/Chic, A/Duck/Ukraine/1/63 (H3N8): A/Duck, A/WSN-FLAG (H1N1):
A/WSN(F), B/Memphis/12/97 (Yamagata-lineage): B/Mem, B/Wisconsin/01/2010 (Yamagata-lineage): B/Wis, and B/Brisbane/60/2008 (Victoria-lineage): B/Bris. A/
NIBRG is an influenza stain derived from a highly pathogenic H5N1 strain (A/
Vietnam/1194/2004), in which the HA is adapted to a low-pathogenic form. Only
0.1 1 10 100
0 25 50 75 100 125
Concentration (µM) Normalised fraction of infected cells (%)
0.1 1 10 100
0 25 50 75 100 125
Normalised fraction of infected cells (%)
Concentration (µM)
0.1 1 10 100
0 25 50 75 100 125
Normalised fraction of infected cells (%)
Concentration (µM)
7-DMA Peptide R50 Peptide 12B
a
0 5 10 15 20 25
0 100 200 300 400 500
Th-T fluorescence 485 nm (a.u.)
Time (h)
0 10 20
0 250 500 750 1000
Time (h) Average RH (nm)
0 1 2 3 4 5 6 7 0
20 40 60 80 100 120
Time (h)
Soluble fraction
Light scattering (10 µM) Amyloid-specic dye (10 µM) TEM (10 µM) Solubility (10 µM)
800 nm
b c
e
d
0.1 1 10 100
0 25 50 75 100 125
Normalised cell viability (%)
0.1 1 10 100
0 25 50 75 100 125
Normalised cell viability (%)
Concentration (µM)
7-DMA Peptide R50 Peptide 12B
0.1 1 10 100 1000
0 25 50 75 100 125
Normalised cell viability (%)
Concentration (µM) Concentration (µM)
f g
h i j
Fig. 6 Peptide R50 organizes into amyloid structures and interferes with ZIKV replication. aHydrodynamic radius (RH) calculated from the regularizationfit of DLS data of peptide R50 (10µM) maturation over time.bTh-T emission signal of 10µM peptide R50 at 485 nm after excitation at 440 nm as a function of time (n=3 independent experiments).cTEM image of 10µM peptide R50 after 6 h incubation.dSoluble fraction of peptide R50 (10 µM) determined after ultracentrifugation (250,000 ×gfor 30 min), measured over time, and mean ± SD is shown (n=3 independent experiments). Dose- dependent effect of 7-DMA (e), peptide R50 (f), and peptide 12B (g) on the fraction of ZIKV-infected cells following 48 h of infection with a MOI of 0.1.
Data are normalized to DMSO-treated, infected cells (100% infected cells) and mean values ± SD are shown (n=3 independent experiments).
Representative images are shown in Supplementary Fig. 10. Dose-dependent effect of 7-DMA (h), peptide R50 (i), and peptide 12B (j) on cell viability of Vero E6 cells without ZIKV infection. Data are normalized to DMSO-treated, noninfected cells (100% cell viability) and the mean values ± SD are shown (n=3 independent experiments).
